Implement MoneyFormatter with Intl support

[henriquemoody]

Summary

Create a MoneyFormatter that uses PHP's Intl extension (NumberFormatter) for currency formatting with a clean, minimal API.

Proposed Design

The formatter should leverage NumberFormatter for locale-aware currency formatting while keeping a simple interface:

use Respect\StringFormatter\MoneyFormatter;

// Simple case - uses default locale
$formatter = new MoneyFormatter('USD');
echo $formatter->format('1234.56');
// Output: "$1,234.56"

// With custom locale
$formatter = new MoneyFormatter('BRL', 'pt_BR');
echo $formatter->format('1234.56');
// Output: "R$ 1.234,56"

// With custom precision
$formatter = new MoneyFormatter('JPY', precision: 0);
echo $formatter->format('1234.56');
// Output: "¥1,235"

Constructor Signature

public function __construct(
    string $currency,
    string|null $locale = null,
    int|null $precision = null
)

• $currency: ISO 4217 currency code (e.g., 'USD', 'EUR', 'BRL')
• $locale: Optional locale for formatting (defaults to system locale)
• $precision: Optional decimal precision (defaults to currency's default)

Key Features

1. Intl integration: Use NumberFormatter::CURRENCY for all formatting
2. Locale-aware: Respect locale-specific formatting (symbol position, separators)
3. Precision control: Override currency defaults when needed
4. Error handling: Validate currency codes and throw InvalidFormatterException
5. Currency validation: Use NumberFormatter to verify valid currency codes

Questions/Decisions Needed

• Should we support currency code-only output (no symbol)?
• Should we support negative numbers with specific formatting?
• Should we format as accounting style for negatives?

[alganet]

It would be weird for an user of Respect\Validation that wants to format a template to use currency formatting to need both intl and sokil/iso-codes at the same time.

We should probably take a step back and rethink how we manage graceful fallbacks in internationalization dependencies, and work on a strategy that works for all our package offerings.

[henriquemoody]

I agree, we need to rethink that aspect. I think we can use them for different scenarios.

Unfortunately, we cannot choose one or another, because they provide different things. Intl doesn't have support for country subdivisions, for example, we still need sokil/iso-codes for that.

[alganet]

CLDR does offer subdivisions and also currency data. It also offers unit information (metric, imperial), and all sorts of things (emoji descriptions, keyboard mappings, etc).

But I don't think it's complete for currencies (it doesn't have symbol-narrow for R$ for example). Also, no postal codes or TLD-country association.

From my initial research, there is no silver bullet database that offers everything.

Also, from a developer's perspective, I almost never want to fully internationalize to all countries. What I want is english and a couple of other languages. Full I18N+L10N is a rare need (but worth pursuing). It would be nice to have a library that supports granular imports:

    dependencies:
        - regional-data-core
        - regional-data-locale-pt-BR
        - regional-data-locale-en-US

Each granule would only contain data for the locale it needs (the intersection of a language and a country).

We could then rely on such regional-data-core and leave the database management to the library user.

Unfortunatelly, there is nothing like that around, except in distros like debian that do offer this kind of granularity.

[alganet]

I was wrong, CLDR does offer R$

https://github.com/unicode-org/cldr/blob/410fa03400aab2c32e90278af9310a93bed2b3ce/common/main/root.xml#L4583

I was looking in pt_BR.xml but it's in root.xml.

Subdivisions:

https://github.com/unicode-org/cldr/blob/410fa03400aab2c32e90278af9310a93bed2b3ce/tools/cldr-code/src/main/resources/org/unicode/cldr/util/data/external/subdivisionData.txt#L533

Unit data:

https://github.com/unicode-org/cldr/blob/410fa03400aab2c32e90278af9310a93bed2b3ce/common/testData/units/unitsTest.txt#L43

It's fairly complete (except for postal codes and TLD info).

[henriquemoody]

CLDR is indeed amazing.

There's a PHP package (or packages) that encapsulate it all. Ideally, we would have a collection of packages that could be installed according to specific needs, for example composer require cldr/languages, composer require cldr/countries, composer require cldr/currencies.

[alganet]

I actually want both. I want to be able to do:

 - dependencies: [ locale-data/currencies, locale-data/en, locale-data/es ] # A financial app in english and spanish
 - dependencies: [ locale-data/en-us, locale-data/en-ca, locale-data/fr-ca locale-data/es-mx ] # An app specific for north americans
 - ...

Specific slices with some overlap (for example, currencies would have lots of currency data but no translations).

By the way I just reserved github.com/locale-data 🐼 I've been digging this hole for the past few days, and it's deep. I think I can reasonably make a solid package collection in 3 months or so. Until then, if there's a package that offers less granular data, I would gladly accept it.